Re: java.util.Timer not handling negative System.currentTimeMillis()

[Andrew Haley <aph at redhat dot com>]

Martin Egholm Nielsen writes:
 > >  > I just encountered a funny (initially strange) problem arising because 
 > >  > my embedded RT-Clock was set to the year 1909.
 > >  > Hence System.currentTimeMillis() returns a negative number, and this is 
 > >  > used internally in java.util.Timer#schedule(), and throws an unchecked 
 > >  > exception - aarghh...
 > >  > 
 > >  > I know this is an obscure problem, but is there is requirement that we 
 > >  > are only "allowed" to run gcj on targets with clocks > 1970.
 > >  > One could change the Timer code to handle this situation, or just ignore it.
 > >  > 
 > >  > For now I will just make sure the time is not less than 1970 and larger 
 > >  > than some other limit overflowing the long.
 > >  > 
 > >  > The problem arises because my target's RTC comes with an undefinded 
 > >  > initial time - hence anything is possible :-)
 > > 
 > > This is a tricky one, because even if we fix gcj to allow this I
 > > suspect some user code will be confused.  It is perhaps not
 > > unreasonable for a programmer to assume that the "current time" during
 > > a program's execution will not predate her birth!
 > > 
 > > You need to make sure that your RTC is initialized to some sane value.
 >
 > Now, that's the hole key issue: "Sane" is not necessarily 1970+ for one 
 > that doesn't bother about the absolute time, but just wants to do 
 > something in relative time...

"Sane" in this context just means > 1970.  If you want to work in
relative time, start at 1970.

Andrew.